1
Teleosemantics and Indeterminacy
David Papineau
Department of Philosophy
King's College London
Strand
London WC2R 2LS
UK
d.papineau@kcl.ac.uk
2
Teleosemantics and Indeterminacy
I.
Introduction
The aim of this paper is to defend the teleological
theory of representation against an objection by Jerry
Fodor. I shall argue that previous attempts to answer
this objection fail to recognize the importance of
belief-desire structure for the teleological theory of
representation.
The central thesis of the teleological theory (I shall
take "of mental representation" as read from now on) is
that the representational contents of mental states can
be explained in terms of the biological functions of
those states. More specifically, this theory equates the
contents of informational states (and in particular,
therefore, of beliefs) with those conditions in which
those states are biologically supposed to be present;
and it equates the contents of affective states (and in
particular, therefore, of desires) with the conditions
which those states are biologically supposed to produce.
(For versions of this theory, see Dennett [4, ch. 9], [6,
ch. 8]; Fodor [7]; Millikan [12], [14]; Papineau [19],
[20], [22]; McGinn [10, ch. 2].)
Defenders of the teleological theory standardly cash
out their references to biological purposes
etiologically, in terms of past processes of natural
selection. According to the etiological account of
biological function, an item X has a biological purpose Y
if and only if X is now present because previous versions
of X were selected in virtue of doing Y. (Cf. Wright
[30]; Millikan [14, ch. 1]CHECK; Neander [16], [17].)
The paradigm case of such selection is inter-generational
Darwinian selection of genetic traits. But many
advocates of the teleological theory (though not
Millikan) also take representational functions to derive
from non-genetic selection of mental states during
individual learning.
II.
Fodor's Objection
The teleological theory developed in response to
"indicator" theories of content (Dretske [6], Stampe
[27]). Theories of this kind equate the contents of
mental states with the conditions that those states covary with. The difficulty facing such theories is that
not all co-variation seems to be representational (Fodor
[7]). My temporal beliefs co-vary with the position of
the hands on my watch, but represent something else,
3
namely, the local time, which will be different if my
watch is faulty.
The attraction of the teleological theory is that it
seems to impose extra constraints on which kinds of covariation count as representational. It is not enough
that a mental state actually co-vary with such-and-such a
condition. The teleological theory requires in addition
that the state is biologically supposed so to co-vary.
Fodor [8, ch. 3] argues that this advantage is
illusory. He considers the state that is produced in a
frog's brain when its eye is stimulated by a nearby small
black moving object. At first sight, says Fodor, it may
seem as if the teleological theory will ascribe the
content fly to this state, in line with pre-theoretical
intuition. For the frog state has presumably been
favoured by natural selection because of the advantageous
consequences it produced on those occasions it was copresent with flies. However, continues Fodor, this isn't
the only answer the teleological theory authorizes. For,
given that most small black things in the frog's
evolutionary history have been flies, isn't it equally
true that the frog state has been selected because of the
advantageous consequences it produced when co-present
with small black things? So this seems to imply that the
teleological theory should ascribe the content small
black thing to the frog's brain state, as much as the
content fly. So Fodor argues the teleological theory is
unable to assign determinate content to states like the
frog's brain state after all.
In addition to his example, Fodor offers a general
diagnosis of the difficulty facing the teleological
theory. He ascribes the difficulty to the
"extensionality" of natural selection. According to
Fodor, it matters to natural selection which situations a
given trait has advantageous results in; but it doesn't
matter to natural selection how that class of situations
is described. So, given that "fly" and "small black
moving thing" have been more or less co-extensional in
frog history, it is unsurprising that the teleological
theory is unable to decide between them as readings of
the frog brain state.
I think that, while Fodor is right that frog-type
examples present a difficulty for the teleological
theory, he is wrong in his general diagnosis of the
difficulty. There seems to me no good reason to think
that natural selection is "extensional" in the way Fodor
suggests. On the contrary, biologists are not so much
interested in when a trait is advantageous, as why it is
advantageous, and for that reason restrict themselves to
causally relevant descriptions of biological situations.
4
(Cf. Millikan, [15, p. xx]CHECK; Neander [18, pp. v,
xiii-xiv]CHECK.) If this were not so, it wouldn't only
be "fly" and "small black thing" that were candidate
descriptions of the frog state's function, but also any
grue-like characteristic that just happened accidentally
to have been present on all past occasions when the
frog's state proved advantageous. Maybe there are
underlying metaphysical puzzles about the fact that "fly"
and "small black thing" are causally relevant
descriptions, while grue-like descriptions are not; but
since nobody would say that these metaphysical puzzles
invalidate biological analysis in general, it would be
unreasonable to argue that they are an obstacle to a
biological analysis of representation in particular.
(Cf. Sterelny [28, pp. 00-00]CHECK.)
Still, the frog problem does point to a general
indeterminacy in selection-based function ascriptions,
even if it is not the indeterminacy that Fodor
identifies. The real source of the frog-problem is not
that natural selection is "extensional", but rather that
the ascription of biological functions still seems
indeterminate even if we restrict ourselves to causally
relevant descriptions. To borrow an example of Karen
Neander's [18, pp. 00-00]CHECK, imagine that a species of
highland antelope has some distinctive trait T which has
been selected because it (a) alters the antelope's
haemoglobin structure (b) increases oxygen uptake (c)
enables the antelope to live on higher ground (d) gives
it access to a plentiful food supply (e) increases
reproductive success. Which of this "concertina" of
effects is T's function? There seems no obvious answer.
After all, by hypothesis it is because T has produced all
the effects in the concertina on various past occasions
that it has been selected. So the etiological notion of
function seems to provide no basis for identifying any
one of the effects as T's function.
We can view the frog problem as the consequence of
this general indeterminacy for the teleological theory of
representation. The frog's brain state enables the frog
(i) to detect small black things, and thereby (ii) to
detect flies. But since the frog's state was selected in
virtue of its doing both these things, the etiological
notion of function does not discriminate between them.
III.
Millikan's Response
I shall now proceed by considering in turn the responses
to Fodor made by Ruth Millikan (section III) and Karen
Neander (sections IV-VI). I shall argue that, while each
contains a good idea, they both founder because they pay
insufficient attention to the significance of beliefdesire structure. I shall then give my own response to
5
Fodor (sections VII-X), which will appeal to these two
good ideas but will also focus specifically on the
representational purposes of belief and desires.
Ruth Millikan argues that Fodor's problem arises
because he is focusing on the causes of the frog's brain
state, rather than the way in which the frog uses the
state in directing actions. The purpose of the state
does not depend on which condition is supposed to cause
it, argues Millikan, but rather on which condition will
enable the use of the state to fulfil some further
biological function. (Millikan [14, esp. chs 4 and 11],
[15].)
The frog's state makes the frog stick its tongue out.
What ensures that sticking its tongue out will have
biologically advantageous results for the frog? Not just
a small black thing, for this certainly won't lead to
advantageous results, if it is a speck of dust rather
than a fly. Nor is it exactly the presence of a fly,
since any other edible bug will suit the frog equally
well, and an inedible fly won't. So Millikan argues that
the content of the frog state is something like food,
rather than small black moving thing, or even fly [15, p.
xxvi]CHECK).
I agree entirely with Millikan that we need to focus
on results, rather than causes. If a biological trait
has a function, this is because it produces advantageous
effects. In the case of an informational state, it will
produce advantageous effects by being present in suchand-such circumstances. The teleological theory should
therefore identify the content of an informational state
with the circumstances in which it so leads to
advantageous effects.
Where Millikan goes wrong, however, is equating
"advantageous effects" here with any biologically
advantageous effects. As I shall argue in sections VIIX, the teleological theory will only work if we equate
advantageous effects specifically with effects that
satisfy desires, rather than with any biologically
advantageous effects.
Without this modification, Millikan's suggestion runs
into the general "concertina" problem raised at the end
of the last section. She says that the frog state
represents food, rather than fly, or small black moving
thing, because it is only when food is present that the
frog action has advantageous effects. But why stop
there? If the food triggers an allergic reaction, or in
any other way fails to foster the frog's health, then no
biological advantage accrues. So why not take the frog
state to represent health-preserver, rather than food?
6
By the same coin, why not take the frog state to
represent reproduction-enhancer? Eating the fly won't
have any advantageous effects which contribute natural
selection, on those occasions where, say, the fly
contains some substance that sterilizes the frog, or for
any other reason does not help the frog to find a mate
and reproduce. (Cf. Goode and Griffith, [9, p. 00]CHECK,
Neander [18, pp. xvi-xviii]CHECK.)
IV.
Neander's Response
Karen Neander tries a different tack. She starts with
the general concertina problem, as exemplified in her own
example of the highland antelope. Her response to this
problem is that all the effects of the antelope's trait T
are indeed genuine functions. Her reasoning is that,
when everything works as it is biologically supposed to,
then the antelope's trait does produce all these effects.
I agree with Neander that this is the right thing to say.
The trait T has the whole concertina of effects as its
functions. After all, the past selection of T occurred
precisely when all these effects were produced, rather
than any subset.
However, Neander then points out, if we focus on the
notion of malfunction, rather than of function, then it
is possible to identify one of these effects as a
function which is specific to the antelope's trait T.
The point is that not all the effects in question are
ones whose absence means that T is malfunctioning.
Suppose that there is in fact no increased oxygen uptake
in some particular member of the antelope species. Does
this mean that T is not doing its job? Not necessarily,
because increased oxygen uptake also depends on other
traits, such as efficient lungs, and so the lack of
increased oxygen uptake may be due to the lungs
malfunctioning instead. In such a case, where the
antelope has deficient lungs, say, the trait T will not
perform most of its functions. But it would be wrong on
that account to say that trait T is malfunctioning, since
the absence of these effects is the fault of the
deficient lungs, rather than T.
Of course, if there is no increased oxygen uptake (or
highland habitat, or reproductive success) this may be
because T is malfunctioning. But it could equally be
because other parts of the organism are not doing their
job. So if we ask, "Which absence ensures that T is
malfunctioning?" we get a unique answer, "The absence of
that effect that depends on T alone, namely, altered
haemoglobin structure."
The point of Neander's emphasis on malfunction is that
it gives us a rationale for identifying, among the
7
different effects that have been responsible for a
trait's selection, one particular function that is
specific to that trait. Neander amplifies this idea by
relating it to Robert Cummins' notion of "functional
analysis" [3]. Cummins points out that that the workings
of any complex entity like a biological organism can be
successively decomposed into subsystems. The first
decomposition of an organism might give us the
circulatory system, the respiratory system, the immune
system, the nervous system, and so on; the circulatory
system can then be decomposed into the heart, the blood
vessels, the blood, and so on. At each level the
contribution made by each part to the working of the
whole can be described. Neander argues that the function
specific to any trait is the immediate effect it produces
at the lowest level of description where it appears as an
unanalysed component in the functional analysis. The
idea is that, given any trait T, we should analyse the
system until T features as a basic but not itself
decomposed component, and then note what immediate effect
T produces at this level of decription.
Neander then argues that this general idea provides an
answer to Fodor's indeterminacy challenge to the
teleologial theory of content. More specifically, she
holds that if we apply her general analysis of functions
to the frog, then we can conclude that the frog's state
represents small black things, rather than flies, or
food, or reproduction-enablers. For, if we take the
lowest level at which the frog's "bug-detection" system
is an unanalysed component, argues Neander, then the most
immediate effect produced by this system is detecting
small black things. It is only as a result of doing this
that the system detects flies, or frog food, or
reproduction-enablers. So the content of the states of
this system is small black things.
V.
Frogs are Bad Examples
I think that Neander has the right solution to the
general indeterminacy problem facing all ascriptions of
functions. We can identify the function specific to a
given biological trait by focusing on cases where that
trait malfunctions.
Moreover, as I shall explain later, I think that this
general analysis of functions helps answer Fodor's more
specific objection that teleosemantic ascriptions of
content are indeterminate. However, I do not think that
Neander's own use of her general analysis is the right
way to answer Fodor. My answer to Fodor will use her
general analysis of functions in a rather different way.
8
In a moment I shall explain what is wrong with
Neander's own answer to Fodor. But first I would like to
enter a general complaint about the excessive attention
paid to the frog example in the literature on Fodor's
objection.
Note that our discussion so far has shown that there
are no clear pre-theoretical intuitions about what the
frog's state does represent. Different participants in
the debate favour quite different ascriptions of content
to the frog state. Fodor's initial discussion assumes
that a good theory ought to make this state represent
fly. Millikan is happy with a theory that implies it
represents food. Now Neander defends a theory which has
it representing small black thing.
One obvious consequence is that we are not going to be
able to evaluate theories of representation by noting
what they imply about frogs. In the absence of any pretheoretical agreement, each theory is free to interpret
the frogs in its own way.
Moreover, I suspect that behind this methodological
point there lies a more substantial moral. Namely, that
the reason there are no firm intuitions about the content
of the frog's state is that there isn't really any fact
of the matter about it. My own response to Fodor will
imply that a cognitive state only has a definite content
when it is part of a psychological structure with the
complexity of belief-desire psychology. A corollary is
that the frog's state, which is not embedded in such a
structure (at least as the frog is described in the
philosophical literature)1, does not in fact definitely
represent fly, rather than frog food, or small black
thing.
Some commentators on the teleological theory argue
that its first task is to deal with biologically simple
examples like the frog's state. If the theory can't cope
with the simple representational states of frogs, then
what chance does it have with the far more complicated
representational states of human beings? (Cf. Agar [1,
pp. 1-2].) After all, it must be easier to identify the
functional facts in simple animals like frogs than in
more complicated organisms like human beings. And if
representational facts depend on functional facts, as the
teleological theory argues, then doesn't it follow that
frog representation must be easier to undertstand than
human representation?
But this is too quick. We don't yet know how
representational facts are related to functional facts.
Even if we assume that they are somehow related, the
discussion so far shows that the dependence is not
9
obvious. So it is perfectly possible that the kind of
functional facts required to fix representational facts
display a complexity which the frog lacks. This is what
I shall argue. The frog's traits have perfectly good
functions, but these aren't the kind of functions that
fix determinate representations. Representation only
becomes determinate when we move away from frogs towards
beings with more complicated psychologies.
True, by the time we reach the full complexity of
human beings, the teleological theory faces various
further problems. (For example, it needs to show how the
many sophisticated mental states that humans acquire in
the course of individual development, most of which have
no obvious biological relevance, somehow derive their
representational powers from genetic or non-genetic
selection processes.) However, such further problems are
not my present concern. My aim here is merely to answer
Fodor's indeterminacy objection, and my contention is
that we will not be able to deal with this objection
until we have moved away from frogs to beings with
beliefs and desires.
VI.
Objections to Neander
I still owe my response to Neander. Recall that Neander
takes the frog state to represent small black things, on
the grounds that "detecting small black things" is the
most immediate effect produced by the frog's bug-detector
at the lowest level where it is an unanalyzed component.
Neander's strategy is to identify the functions
specific to given biological traits by noting which of
their effects are most immediate. And she says that
"detecting small black things" is the most immediate
effect of the frog's detection system. My fundamental
objection to this strategy is that "detecting small black
things" doesn't seem to be an effect at all. The
following are surely the effects of the frog's detection
system: the frog sticks its tongue out; it catches a
small black thing; it catches a fly; it gets food; and
so on. If anything, the most immediate effect is
sticking its tongue out. In any case, "detecting suchand-such" does not feature among the effects.
I would say that "detecting X" is a cognate notion to
"representing X". It is the kind of notion that the
teleological theory needs to analyze, not a notion it can
take as given. In order for Neander to solve the frog
problem, she ought to start with the structure of effects
produced by the frog's detection system, and thereby
reach a conclusion about what it detects. It seems to
beg the question to have "detecting Xs" appearing among
the effects.
10
Couldn't Neander say that the most immediate effect of
the detection system is that the frog sticks-out-itstongue-in-response-to-small-black-things?2 But if we are
going to include the stimulus producing the response as
part of the description of the most immediate effect, why
not say the most immediate effect is sticking-out-itstongue-in-response-to-flies, or sticking-out-its-tonguein-response-to-food, or so on? These all seem
respectable alternative decriptions of the most immediate
effect of the frog's detection system, namely, its
sticking out its tongue.
True, actually catching small black things is a more
immediate effect than catching flies, or getting food.
But this is a matter of the effects which issue from the
frog's state, not of what the state is responding to. It
would seem to be a confusion to infer from this that
sticking-out-its-tongue-in-response-to-small-black-things
must somehow be privileged over sticking-out-its-tonguein-response-to-flies. There is no obvious reason to hold
that the most immediate effect (sticking out its tongue)
must be characterized as a response to the condition
required for the next most immediate effect (small black
things). After all, as I have just observed, it seems to
make perfectly good initial sense to say the frog is
sticking out its tongue in response to flies, or food, or
so on.
I would argue that Neander has taken insufficient note
of Millikan's point that representational content hinges
on how the representation is used, not on what causes it.
In her general discussion of teleology, Neander focuses,
quite rightly, on the effects of biological traits. But
as soon as she turns to representation she shifts to the
question of what is supposed to cause the frog's state
(what it is supposed to detect). From the teleological
point of view, this is to start at the wrong end. The
teleosemantic strategy requires us first to identify
which result the state is supposed to produce, and then
use this to tell us what it is representing. Neander,
however, tries to decide directly what the state is
representing, without consdering what effect the state is
supposed to produce. Perhaps further elaboration could
show how this input-orientated strategy can be made to
work. But such a strategy would have more in common with
older indicator theories of content than with the kind of
teleosemantic theory that Neander is trying to defend
7
An Alternative Solution
Let me now offer my own solution to the indeterminacy
problem. I shall first state this solution briefly, and
then go into details.
11
Consider some animal that has definite desires.
Suppose in particular that it desires to eat a fly. (It
wants to eat a fly, not just any small black thing, nor
any other food -- just as a human being might want to eat
chocolate, not any stuff that looks brown and soft, nor
any other food.) Suppose in addition that this desire,
when combined with some belief S, causes the animal to
shoot out its tongue. If we now assume that the
biological purpose of beliefs is to be present in those
circumstances where the behaviour they prompt will
satisfy the desires they are combining with, then it
follows that the content of the belief S is that a fly is
in tongue-reach, rather than food, or a moving black
thing. For it is only when a fly is in fact within
tongue-reach that the animal's desire for a fly will be
satisfied. If it is responding to food which isn't a
fly, or a small black thing which isn't a fly, then it
won't satsify its desire for a fly by catching it.
This analysis is just a version of Millikan's idea
that we should identify the content of an informational
state with that condition which will ensure that the
behaviour this state generates will have biologically
successful results. But it goes beyond Millikan in
focusing specifically on results which are successful in
the specific sense of satisfying desires, not on any
results which are biologically successful in any sense.
It is because of this restriction that my suggestion
avoids the indeterminacy which remains in Millikan's
approach. While food, health and reproductive success
are all no doubt successful results which are
biologically supposed to follow from the animal's
behaviour, the circumstances required for them to follow
won't count as the content of S, on my suggestion, unless
these are results that the animal desires.
If this account of belief contents is right, then it
will follow that animals which lack a belief-desire
psychology are cognitively too simple to have mental
states with definite contents. The contents of mental
states are only pinned down by their roles in an
interlocking structure of beliefs and desires.
8
Beliefs and Desires
The suggestion in the last section was an instance of a
familiar strategy for fixing belief contents in terms of
desire contents: a belief's content is that condition
which will ensure the satisfaction of whichever desires
it combines with to generate behaviour. (Cf. Ramsey [25,
p. 159], Putnam [24], Appiah [2], Mellor [11], Whyte
[1990], Papineau [21].) Some readers may be curious at
this point why we should do things this way round.
12
Suppose we take it as given that belief contents and
desire contents are interdependent. Even so, why assume
that we should fix beliefs contents in terms of desire
contents? Why not fix desire contents in terms of belief
contents? Or why not fix both in terms of something
else?3
All these different directions of explanation have
been pursued in the literature. But the teleological
approach I am defending here is committed to explaining
belief contents in terms of desire contents. This is
because biological functions are always a matter of
effects -- functions are effects in virtue of which
traits are selected. Beliefs, however, do not have any
effects to call their own. Rather, their biological
purpose is to produce whichever results will fulfil the
purposes of the desires they are acting in concert with.
So the teleological theory will start with the results
which desires are biologically supposed to produce, and
which it will count as the satisfaction conditions of
those desires. And then it will then explain the
contents of beliefs derivatively, by saying that the
purpose of beliefs is to generate actions which will
produce desired results in such-and-such conditions,
which conditions will therefore count as the truth
conditions of those beliefs.4
So I am committed to explaining belief contents in
terms of desire contents. Given this, however, it may
seem that I am still open to an obvious variant of
Fodor's objection. For doesn't Fodor's problem arise as
much for desires as beliefs? If we think of the contents
of desires, in line with the teleological theory, as the
results that they are biologically supposed to produce,
then why fix on the content fly, say, for a given desire?
After all, aren't small black thing, food, health,
survival, and reproductive success equally effects that
this desire is biologically supposed to produce? (Cf.
Fodor [8, p. 72].)
If desire contents are as indeterminate as belief
contents, then the strategy of using desres to tie down
beliefs is not going to work. However, I now think we
now have the resources to answer this objection. We can
use Neander's general account of which functions are
specific to which biological traits to identify the
contents of desire states.5
The general idea will be that the function which is
specific to a desire, in Neander's sense, should be
counted as its satisfaction condition. I shall elaborate
this idea in the section after next. But first it will
be useful to make some further remarks about the
13
interlocking biological functions of beliefs and desires
in beings like ourselves.
The function of beliefs is perhaps the more obvious.
Beliefs allow organisms to tailor their behaviour to
circumstances. We can bring this out by comparing beings
like ourselves with a simpler sort of organism (let us
call them "simpletons"). "Simpletons" don't have two
kinds of mental states, informational and motivational,
which interact in generating behaviour. In so far as
they have mental states at all, each is triggered by some
specific stimulus and leads to some specific behaviour.
We are supposing frogs are like this. Any small black
thing triggers a state which makes the frog stick its
tongue out. But human beings are different. Our desires
do not always generate the same behaviour. Which
behaviour issues from a given desire depends on which
beliefs it is acting in concert with. The point of this
dependence of behaviour on belief is that it allows us to
vary our behaviour according to the circumstances. If
the belief does its job properly (that is, it is held in
circumstances which make it true), then the behaviour
chosen will be appropriate for producing the desired
result in those circumstances. So we are able to do much
better than the "simpletons", in that we can pursue our
desires with appropriately flexible responses across the
range of differerent circumstances our beliefs inform us
about.
Let us now focus on the role of desires. The role of
desires in our belief-desire psychology is less often
remarked than that of beliefs, which is a pity, since it
is in some ways less obvious. From the point of view of
evolution, it is not clear why we have a range of
different desires at all. Why don't we just choose
whichever behaviour will maximize reproductive success?
To see the point of this question, compare ourselves with
a type of organism I shall call a "super-brain". "Superbrains" are like us, and unlike simpletons, in having
beliefs, but their beliefs always inform the singleminded pursuit of reproductive success. Super-brains
always choose the action which they believe will maximize
reproductive success. Given sufficiently streamlined
brains, these super-brains would be biologically ideal
beings. However, we human beings are different,
presumably because it would be too difficult, given our
limited cognitive capacities (not to mention the limited
capacities of our evolutionary ancestors), for us
reliably to work out in real time which actions will
maximize reproductive success. So instead natural
selection has given us states (innate and acquired
desires) which make us use our beliefs to select actions
which will produce certain intermediate results, like
food and warmth and coca-cola. These results correlate
14
reasonably well with reproductive success, which is why
we have evolved to seek them. But even when we can see
that they conflict with reproductive success, we seek
them anyway, which is the difference between us and the
"super-brains".
The difference between us and the super-brains may not
be fully clear, Won't super-brains also have
intermediate desires? For example, mightn't a superbrain judge that the best current means to its
reproductive success is to eat some ice-cream, and for
this reason desire some ice-cream? But while a superbrain might pursue ice-cream as a means, it won't desire
it as we do. For by hypothesis the super-brain will
cease to pursue the ice-cream as soon as it comes to
believe that it isn't good for its health, or is
otherwise deleterious to reproductive success. We aren't
like that. Even though my desire for ice-cream is
presumably environmentally acquired, rather than
genetically innate, it will still persist even after I
discover ice-cream only makes me fat. Life would be much
simpler if all our desires for intermediate results
simply dissolved as soon as we discover that they
interfere with more ultimate biological goals.6
9
Psychological Realism
Some readers may be worried about the status of the last
section's relatively detailed empirical claims about the
structure and biological function of human belief-desire
psychology. A full response to this worry is clearly not
possible here, but a couple of brief comments may be
helpful. (For further discussion see Papineau [22],
[23].)
Consider the "simpletons" and the "super-brains"
again. I have taken it that we humans are different from
both. I have assumed that we are different from the
simpletons, in that our behaviour varies depending on
what our beliefs indicate to be the best means to certain
ends. And I have assumed that we are different from the
super-brains, in that the ends we pursue are the
intermediate results we desire, and not just reproductive
success.
So far I have offered no evidence for these claims
about human beings. I have simply presented them as
truisms of everyday psychology. But we do not have to
rest with the fact that everyday psychology says we are
different from the simpletons and the super-brains. For
there is any amount of direct behavioural evidence to
show we are so different. After all, we don't behave
like simpletons or super-brains. As soon as I learn that
the pub has no beer, the arrival of 7 pm ceases to
15
stimulate me to go there. So I'm not a simpleton. And
even after I discover that drinking beer only hinders my
reproductive success, I continue drinking it. So I'm not
a super-brain.
Clearly this kind of evidence can be multiplied
indefinitely. The structure of our behaviour would be
quite different if it were not directed by beliefs and
desires. So we have every reason to accept that we
really do have the structure of beliefs and desires
posited by everyday psychology.7
Moreover, if we do have this structure, then
presumably it is a matter of biological design. It would
be very odd to accept that humans do indeed have a
structure of beliefs and desires, yet deny that this
structure has evolved because it enables humans to tailor
their behaviour to current circumstances and thereby
achieve outcomes which are correlated with (but distinct
from) reproductive success. We should not of course
believe every fanciful "just-so story" which is offered
in explanation of any biological trait. But in the case
of human belief-desire psychology it is hard to see how
there could be any other possible explanation apart from
evolutionary design. (Cf. Millikan [13, pp. 292-4].)
10
The Contents of Desires
The task I still face is to show how the teleological
theory can pin down the content of desires in the face of
Fodor's indeterminacy objection. My strategy is to take
Neander's analysis of which functions are specfic to
which traits, and then apply it, not as she does, but
rather within the conext of belief-desire psychology, to
pin down which of the results produced by a desire is its
specific function. So my suggestion is that the
teleological theory should identify the satisfaction
condition of a desire as that result which is the
desire's specific function.
It will be helpful to switch examples slightly at this
point, and consider now the desire for food in some
organism with a belief-desire psychology. My task is to
show that getting food into your stomach is the specific
function of this desire, in Neander's sense, in the face
of Fodor's complaint is that this desire is equally
supposed to produce a number of other results.
Consider a typical case where your desire for food
leads you to put the contents of your spoon into your
mouth. If all goes well, your desire has the following
results: you move your right arm, your spoon enters your
mouth, your stomach receives food, the food is digested,
your health is preserved, you survive, you reproduce.
16
Fodor would point out that the biological purpose of the
desire state is to produce just this kind of "concertina"
of results. So what entitles the teleological theory to
pick out just one -- the stomach's reception of food -as the desire's satisfaction condition?
As a preliminary step, it will be helpful to split
this into two problems. First, what shows that the
initial results in the above sequence (you move your arm,
the spoon enters your mouth) are not the satisfaction
condition of the desire? Second, what shows that the
later results (the food is digested, your health is
preserved, you survive, you reproduce) are not the
satisfaction condition of the desire?
The answer to the first problem is obvious enough.
The initial stages in the concertina of results depend as
much on the beliefs behind your behaviour as on the
desire itself. (If you didn't believe that there is food
in your spoon, your desire for food wouldn't make you put
the spoon in your mouth. And if you believed instead
that there is food in the cupboard across the room, then
the desire would make you walk across the room.) So if
we want a result that attaches to the desire as such, and
not just to the conjunction of the desire with specific
beliefs, then we need to look further along the
concertina, to those results which the desire is always
supposed to produce, when things work as they are
biologically supposed to.
We can amplify this answer to our first problem by
contrasting ourselves with the "simpletons". The
cognitive states of simpletons are designed always to
trigger exactly the same sequence of results (tongue
shoots out, black thing caught, . . .) But our cognitive
states work differently. Which sequence of effects is
produced by any given cognitive state depends on which
other cognitive states it interacts with. In particular,
the initial effects produced by any desire will depend on
which beliefs it interacts with, since different beliefs
will indicate different behaviours as the means to the
desire's satisfaction. So if we want to identify effects
which it is the function of the desire to produce, we
need to go far enough along the concertina to reach
results which do not depend on which beliefs the desire
happens to be interacting with.
This now brings us to the second problem. Even if we
can dismiss the initial results in the concertina, we are
still left with the many further results the desire is
always supposed to produce (the stomach receives food,
the food is digested, your health is preserved, you
survive, you reproduce). It is here that I think we
should appeal to Neander's analysis. We can agree that
17
the desire for food not only has the function of putting
food in your stomach, but also the further functions of
allowing digestion, forstering health, and so on. For
presumably it was when this desire caused this whole
sequence of results that selection preserved it.
However, if we apply Neander's analysis, then we see that
the function of acquiring food is nevertheless specific
to this desire. For the non-fulfilment of the further
functions, like digestion and reproduction, doesn't show
that this desire is malfunctioning, since these further
functions depend not just on the desire doing its job,
but also on other traits, like the digestive and
reproductive systems, doing their jobs too. So if the
teleological theory equates the cosntents of desires with
those effects which it is their specific function to
produce, then it implies the wanted conclusion that in
this case the desire is for food.
We can think of the role of desires as follows. When
a desire is activated, it combines with current beliefs
(there is some food in the kitchen, say) to generate
behaviour (going to the kitchen) which will produce some
specific result (food) if those beliefs are true. SO
this specific result (food) is the job of the
psychological system activated by the desire. Further
effects (digestion, health, survival, reproduction)
depend on other systems as well (the digestive system,
reproductive system, and so on). If these further
effects fail to ensue, then something has gone wrong,
biologically speaking. But this won't be a failure of
the psychological system activated by the desire, as long
as this system has done its specific job of getting food.
Note how this explains why the knowledge that some
desired result is biologically disadvantageous does not
necessarily stop us pursuing that result. (Knowing that
beer hinders my reproductive success, say, does not stop
me drinking beer.) The reason is that is not usually the
specific function of the psychological system which
directs our behaviour to ensure ultimate biological
success, but rather to achieve intermediate results,
namely, the satisfaction conditions of desires.
It might be helpful to compare ourselves to the
"super-brains" once more. Our difference from the superbrains brings out the sense in which the specific
function of desires, in Neander's sense, is normally to
produce intermediate goals. As we have seen, there is
not necessarily a psychological malfunction in beings
like us if reproductive success is not achieved, since
the psychological system activated by our desires will
have done its job if we get food, say, even if
reproductive success does not follow. But with the
"super-brains" it is indeed a failing of their psychology
18
if the ultimate goal of maximizing reproductive success
is not achieved. For the job of the super-brains'
psychological system is precisely to work out which
actions will yield maximum reproductive success, and so
this system has failed if it gets this wrong.8 So the
difference between us and the super-brains is that our
psychology, unlike theirs, often has the specific
function of producing some intermediate result, like food
or warmth or coca-cola, depending on which desires are
activated.
11. Conclusion
I have argued that the teleological theory should answer
Fodor's objection by appealing to the structure of
belief-desire psychology. Within this context, it can
equate a belief's content with that condition which will
ensure the satisfaction of desires. The challenge of
indeterminacy then shifts to the satisfaction condition
of desires. This challenge can be answered by appealing
to Neander's analysis of specific functions. This
analysis implies that desires in beings like us have the
specific function of producing specific results, as is
illustrated by the difference between us and the superbrains. The teleological theory can therefore equate the
satisfaction conditions of desires with those specific
results.
_____________________
I would like to thank Tim Crane, Jerry Fodor, Peter
Goldie, Keith Hossack, Michael Martin, Ruth Millikan,
Karen Neander, Lucy O'Brien, Georges Rey, Gabriel Segal,
Barry Smith, Scott Sturgeon, Bernhard Weiss and two
anonymous readers for the Australasian Journal of
Philosophy for commenting on earlier versions of this
paper.
19
Bibliography
1. N. Agar, "What Do Frogs Really Believe?", Australian
Journal of Philosophy 71 (1993), pp. 00-00.
2. A. Appiah, "Truth Conditions: A Causal Theory", in J.
Butterfield (ed.) Language, Mind and Logic (Cambridge:
Cambridge University Press, 1986).
3. R. Cummins, "Functional Analysis", Journal of
Philosophy 72 (1975), pp. 741-765.
4. D. Dennett, Content and Consciousness (London:
Routledge and Kegan Paul, 1969).
5. D. Dennett, The Intentional Stance (Cambridge, Mass:
MIT Press, 1987).
6. F. Dretske, Knowledge and the Flow of Information
(Oxford: Basil Blackwell, 1981).
7. J. Fodor, "Semantics, Wisconsin Style", Synthese 59
(1984), pp. 231-250.
8. J. Fodor, A Theory of Content, (Cambridge, Mass: MIT
Press, 1990).
9. R. Goode and P. Griffiths, "The Misuse of Sober's
Selection of/Selection for Distinction", Biology and
Philosophy 10 (1995), pp. 99-108.
10. C. McGinn, Mental Content, (Oxford: Basil Blackwell,
1989).
11. D.H. Mellor, "I and Now", Proceedings of the
Aristotelian Society 89 (1988), pp. 00-00.
12. R. Millikan, Language, Thought, and other Biological
Categories, (Cambridge, Mass: MIT Press, 1984).
13. R. Millikan, "Biosemantics", Journal of Philosophy,
86 (1989), pp. 00-00.
14. R. Millikan, White Queen Psychology and other Essays
for Alice, (Cambridge, Mass: MIT Press, 1993)
15. R. Millikan, "Speaking up for Darwin" in G. Rey and
B. Loewer (eds) Meaning and Mind: Fodor and his Critics
(Oxford: Basil Blackwell, 1993), pp. 00-00.
16. K. Neander, "Functions as Selected Effects: The
Conceptual Analyst's Defence", Philosophy of Science 58
(1991), pp. 168-184.
20
17. K. Neander, "The Teleological Notion of a Function",
Australasian Journal of Philosophy 69 (1991), pp. 454468.
18. K. Neander, "Misrepresenting and Malfunctioning"
Philosophical Studies 0 (199?), pp. 00-00.
19. D. Papineau, "Representation and Explanation",
Philosophy of Science 51 (1984), pp. 00-00.
20. D. Papineau, Reality and Representation (Oxford:
Basil Blackwell, 1987).
21. D. Papineau, "Truth and Teleology", in D. Knowles
(ed.) Explanation (Cambridge: Cambridge University Press,
1991). pp. 00-00.
22. D. Papineau, Philosophical Naturalism (Oxford: Basil
Blackwell, 1993).
23. D. Papineau, "Reply to Godfrey-Smith, Symposium on
Philosophical Naturalism", Philosophy and
Phenomenological Research, forthcoming.
24. H. Putnam, "Reference and Understanding", in his
Meaning and the Moral Sciences (London: Routledge and
Kegan Paul, 1978).
25. F. Ramsey, "Facts and Propositions", Aristotelian
Society Supplementary Volume 7 (1927), pp. 00-00.
26. R. Stalknaker, Inquiry (Cambridge, Mass: MIT Press,
1984).
27. D. Stampe, "Towards a Causal Theory of Linguistic
Representation", in P. French, T. Ueling and H. Wettstein
(eds) Midwest Studies in Philosophy 2 (1977) pp. 00-00.
28. K. Sterelny, The Representational Theory of Mind,
(Oxford: Basil Blackwell, 1990).
29. J. Whyte, "Success Semantics', Analysis 50 (1990),
pp. 149-157.
30. L. Wright, "Functions", Philosophical Review 82
(1973), pp.139-168.
It is a moot point whether frogs are as simple as
philosophers suppose. The standard philosophical story
is that frogs will go on sticking out their tongues
indefinitely, sphex-like, if prompted with small dark
moving things. But I have heard it said that frogs are
not really sphex-like, but only seem so in laboratories
1
21
where they are fed bee-bees or other small black things
which fail to assuage their hunger. On this story, frogs
who are fed enough flies will stop sticking out their
tongue until they are hungry again. In the interests of
expository simplicity, however, I shall continue with the
standard assumption that the frog's behaviour is a
invariable reflex unaffected by the frog's current
digestive state.
Neander suggested this possibility to me in
correspondence.
2
Something else will be needed, if want to fix both
belief and desire contents, for we cannot solve for two
unknowns (belief contents and desire contents) with one
equation (the truth of beliefs ensures the satisfaction
of desires). Cf. Stalnaker [26, esp. pp. 15-18],
Papineau [19, p. 555].
3
Perhaps some beliefs also have a different kind of
purpose. Consider the belief that some fierce animal is
closer than it is, or, again, the belief that you will
emerge unscathed from some imminent trial of violence.
Perhaps these beliefs serve an extra biological function,
in addition to aiding the satisfaction of desires -namely, the function of making us act in ways which will
protect our biological needs, even at the cost of
frustrating our desires. For example, the fierce-animal
belief might stop you satisfying your curiosity and make
you flee without delay; and the immunity-from-injury
belief might stop you avoiding dangers and make you fight
without fear.) However, if there are such cases, the
teleological theory of representation should ignore these
extra functions, and analyze the truth-conditional
contents of beliefs specifically in terms of their more
standard function of satisfying desires. For note that
the conditions required for beliefs to serve special
need-protecting functions will standardly not be part of
those beliefs' truth conditions. For example, the belief
"the animal is close" serves its special function when
the animal is in fact further away; similarly, the "I
won't be injured" belief does its special work precisely
when a real injury is likely. For further discussion of
this issue, see Papineau [22, ch. 3.4].
4
Let me reiterate the distinction between Neander's
general account of biological functions, which I endorsed
in section 4, and her own application of this idea to the
Fodor's problem, which I criticized in section 6. I want
to use Neander's general account of functions, but I
shall apply it to Fodor's problem in my own way.
6
Given that the overall function of our cognitive
system is to enhance reproductive success, it is sub5
22
optimal that our desires should persist in the face of
evidence that they are unhealthy. But this is the cost
of having a system which has been designed to calculate
the best means to certain proximate results in real time.
It is true that those results will themselves have been
picked out because of their association with reproductive
success in our long-term evolutionary and individual
experience. But since the whole point of having desires
for these results is to avoid our having to consider
their causal relevance to reproductive success every time
we make a decision, it is inevitable that new information
about this relevance will have no immediate effect on our
decisions.
Whether beliefs and desires have real compositional
structure is a further question, which lies at the centre
of the current debate between traditional and
connectionist programmes in artificial intelligence.
However, while this question is relevant to the details
of the teleological theory, it is not crucial to the
topics discussed of this paper. (Cf. Papineau [23].)
7
A complication. One reason a super-brain might make a
mistake about what will maximize reproductive success
would be that its digestive system isn't working, when it
thinks it is. In this case, the failure to achieve
reproductive success would in a sense be due to a failing
of the digestive system too. But note that even here its
not due to the failure of the digestive system alone.
For the super-brain's psychological system has also
failed, in mistakenly assuming that the faulty digestive
system is working, when it isn't.
8